I have a large number of clinically-obtained MR datasets (de-identified of course), many of which have only T1, T2 and FLAIR sequences, but no T1 volume acquisition. I would like to generate automated slice-by-slice segmentation in order to calculate an approximate cerebral white matter volume. Any suggestions for existing software that can do this?
Dear Andrea Bozoki,
I found an interesting article and sent you.
http://www.isita-org.com/jass/Contents/2009%20vol87/PDF/On-Line_bassa/JASs2009_05_Keller.pdf
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4299953/
Regards, Shafagat
Hi Andrea,
Nice to see you in e space. Hope you are well.
I have used Osirix or Horos as a DICOM image processing tool, but there are problems with accuracy when using thick (>2 mm) 2D images and attempting to segment volumes. See attached. This really is problematic with clinical slice thickness of 4 or 5 mm and worsened with interslice gaps. The only surrogate I have used, and it isn't great is the tracing of structures on the 2D images and factoring in the slice thickness for volume estimates. They are pretty rough. Are you going for hippocampal volumes? Combining rough approximations from sagittal and axial 2D data sets and reconciling them to create an average might be more accurate and reproducible. If you have a third plane it could be added too. Obviously none of this is automated.
Sorry to not have a more satisfying answer.
Best regards,
Mark
Mark, how nice to hear from you! I'm not trying to do anything as sophisticated as hippocampal volume from 2D images, only a basic grey-white-csf distinction. The clinical scans are, as you say, 4ish mm thick, but thankfully no gaps. The goal is straightforward: to obtain an estimate of total subcortical white matter volume (minus brainstem and cerebellum). In turn, I need this to estimate a % of white matter occupied by white matter hyperintensities for a project (I am using SPM12 and the LST for that).
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India